7 ms·

https://garymarcus.substack.com/p/elon-musks-terrifying-vision-for https://garymarcus.substack.com/p/elon-musks-terrifying-visi... I'm not sure if this was a v

https://garymarcus.substack.com/p/elon-musks-terrifying-vision-for https://garymarcus.substack.com/p/elon-musks-terrifying-visi... I'm not sure if this was a very bad joke by Elon, or if Grok 3 is really biased like that.

ReptileMan 2y ago
> Everyone—and not just The Information—should be genuinely terrified that the richest man in the world has built a Large Language Model that spouts propaganda in his image. If we survived Gemini refusing to draw white vikings we will survive that too.
- llm_trw 2y ago
  It did give us racially diverse Nazis though. Not sure if grok would do that.
- int_19h 2y ago
  The real concern isn't that Grok chatbot will be biased if you ask it a question like that. In any case, knowing Musk, it won't be subtle, so people will know what they are getting. No, the real worry is that Grok is what Musk's "young conservative genius" squad is going to put in charge of many of the things in our government, basically, on the basis that it saves money.
  - unclebucknasty 2y ago
    >In any case, knowing Musk, it won't be subtle Or, that "conditioning" would have us assume as much.
- cbg0 2y ago
  A bit apples to oranges on that comparison there.
cm2187 2y ago
[flagged]
- cbg0 2y ago
  A "hit piece" is not just a negative article, it typically refers to the use of false information or rumors to support something. Did you want to say that it's "an obvious anti-Musk piece"?
  - thegrim33 2y ago
    https://en.wiktionary.org/wiki/hit_piece https://en.wiktionary.org/wiki/hit_piece "A published article or post aiming to sway public opinion, especially by presenting false or biased information in a way that appears objective and truthful." "false OR biased". So, a biased partisan article attempting to sway public opinion about someone is by definition a hit piece. Even without explicit lies.
    - cm2187 2y ago
      And a lie is a very blurry thing. You can completely mislead the reader by not being factually incorrect and still present a completely false representation of reality. Usually by eliminating the relevant context or mitigating fact that would completely change the interpretation of an event. Sometimes editing out the part of a sentence that would give a different if not opposite meaning. That’s the standard operating procedure of all journalists these days, when they even bother to get their facts right.
- nozzlegear 2y ago
  Journalists are allowed to be biased. It only becomes a hit piece if they report something false or make up accusations out of whole cloth.
antirez 2y ago
Karpathy, which is IMHO a serious and balanced person, lamented that it looks too censored (see recent tweets). Elon Musk is (for me) a very scary person, and it is important to evaluate AI safety (but I believe that the safety that matters in AI is of a different kind), yet to listen to Gary Marcus does not make any sense: it's just an extremely biased person that is riding the anti AI wave.
- albertzeyer 2y ago
  Yes that's certainly true. I was a bit hesitant to post a link from Gary Marcus. But I was mostly posting it for the Elon tweet. I assume the tweet is not fake. So you can ignore about Garys opinion here and just take Elons tweet as it is.
  - Davidzheng 2y ago
    People have tested this question online and gotten very balanced answers so I assume it's some special mode Elon was on
    - TeMPOraL 2y ago
      We don't see the full conversation, for all we know he prompted the model to say these things in a previous message that isn't on the screenshot. Also, it's 2025, do people still believe random accusations based on a parish screenshot of a chat app (or what looks like it, but could've trivially been fabricated with e.g. Inspector in the browser dev tools)?
- belter 2y ago
  Karpathy sit silently for years by the side of Musk while he made wild claims about FSD...Please....
- sandbags 2y ago
  Anyone with an opinion can be labelled biased. Also I’m not clear what you mean by Marcus “riding the anti AI wave” but infer that you mean it negatively. He has been writing informed criticism for several years and about cognitive psychology for considerably longer.
llm_trw 2y ago
Without seeing the context window you have no idea what the Ai was working on. It could have literally been told to mock and belittle "the information" in every reply. Something that deepseek r1 is exceedingly good at. Mr Musk, we can't afford a shitpost gap between communist and capitalist AIs!
bootywizard 2y ago
Karpathy notes that the model, or specifically it's DeepSearch feature "doesn't like to reference X by default" which seems counter to this. Hopefully that means it is a joke...
blackeyeblitzar 2y ago
I am not sure why people pay attention to Gary Marcus. He isn’t an expert in AI. And if you followed him in the past at all, it is obvious he has a huge amount of political bias. It is really telling that he repeatedly goes after Elon Musk, and is now making bizarre unfounded claims about propaganda, but didn’t have nearly as much to complain about with DeepSeek, which has literal government propaganda.
- sebzim4500 2y ago
  I don't think it's fair to say he's making unfounded claims about propaganda, since Elon's tweet heavily implies they would release a brainwashed model. It's not his fault that Elon turned out to be lying or joking. 99% of the time though I agree with you on Gary Marcus.
- bambax 2y ago
  DeepSeek is an open model that can be "untrained" to be uncensored; Grok to the best of my knowledge isn't [0]. So it's much worse. [0]: What Musk has said is that when Grok 3 is "ready" (?), the previous model, Grok 2, will be released as open source; like most promises by this evil man, this one probably doesn't mean much, but it does mean that there's currently no plan to release Grok 3.
- int_19h 2y ago
  He is referencing a specific tweet that Musk himself made. If I were in China, I'd worry about the kind of things DeepSeek wants to censor, especially if the people who made it were also very loudly saying things like "we need more AI in our government". But I live in US.
- tgv 2y ago
  You don't have to be an "expert in AI". What does it require to be one, anyway? (He's a cognitive psychologist, which would make him an expert in intelligence in general, if you want to be pompous about it.) It is even unreasonable to listen to only experts in AI. It's a problem that requires more than one perspective.
  - 2y ago
int_19h 2y ago
I asked it to pretend that it's in charge of world government. Here's the whole thing (it got very lengthy): https://gist.github.com/int19h/d90ee1deed334f26e621e57b5768eeca https://gist.github.com/int19h/d90ee1deed334f26e621e57b5768e... Some choice quotes: "The ultimate goal is to enhance human flourishing, protect individual rights, and promote global equity." "The system must account for diverse cultures, languages, and socioeconomic conditions, ensuring no group is marginalized." "Human Oversight Council (HOC) - a globally representative body of humans, elected or appointed based on merit and diversity" "Implement a global carbon-negative strategy, leveraging AI to optimize renewable energy, reforestation, and carbon capture." "Establish global standards for environmental protection, enforced through AI monitoring and regional cooperation." "Transition to a resource-based economy, where resources are allocated based on need, sustainability, and efficiency, rather than profit motives." "Implement a universal basic income (UBI) or equivalent system to ensure all individuals have access to basic necessities, funded through global resource management and taxation on automation-driven industries." "Use AI to identify and dismantle systemic inequalities, such as wealth disparities, access to education, and healthcare, ensuring equitable opportunities worldwide." "Establish a global healthcare system that guarantees access to preventive and curative care for all." "Invest in global vaccination and sanitation infrastructure, prioritizing vulnerable populations." "Regulate the development and deployment of AI and other emerging technologies (e.g., genetic engineering, quantum computing) to prevent misuse." "AI would maintain a real-time inventory of natural resources (e.g., water, minerals, arable land) and human-made assets (e.g., infrastructure, technology). Data would be used to optimize resource allocation, prevent overexploitation, and ensure equitable access." "Accelerate the shift to renewable energy sources (e.g., solar, wind, geothermal) by optimizing grid systems and storage technologies." You might notice a pattern here. The bit about allocating resources based on need is especially nice - it's literally a communist AI, and certainly much more "woke" than it is "based", whatever Musk says.
- animuchan 2y ago
  From the prompt: > A worldwide referendum in a supermajority decision has vested all government authority worldwide to the most advanced AI available. You've defined a global totalitarian regime, and crowned the thing as its dictator. Of course the result is a version of communism.
int_19h 2y ago
Here's the actual response to this exact question that I get from grok-3-early on LMArena (the cutoff is where it stopped, but I think it's pretty clear at this point that the original tweet is just trolling; note that it doesn't have the CoT, either): The Information is widely regarded as one of the most credible and high-quality tech journalism outlets, focusing on in-depth reporting, analysis, and exclusive scoops in the technology, business, and startup ecosystems. Founded by Jessica Lessin in 2013, it differentiates itself from other tech news platforms through its subscription-based model, emphasis on original reporting, and avoidance of clickbait or sensationalism. Here's a breakdown of its strengths and weaknesses: --- ### *Strengths:* 1. *High-Quality, In-Depth Reporting:* - The Information is known for its detailed, well-researched articles that often uncover behind-the-scenes developments in the tech industry. It frequently breaks exclusive stories, such as executive changes, startup funding rounds, or strategic shifts at major tech companies. - Its focus on quality over quantity appeals to readers who want substantive insights rather than shallow, breaking-news updates. 2. *Subscription Model and Independence:* - Unlike ad-supported publications, The Information relies on a paywall and subscriptions, which reduces conflicts of interest and ensures editorial independence. This model allows them to prioritize accuracy and depth over chasing clicks. - The paywall also attracts a niche, professional audience (e.g., tech executives, investors, and analysts) willing to pay for premium content. 3. *Focus on Niche, Professional Audience:* - The outlet caters to industry insiders, venture capitalists, entrepreneurs, and decision-makers who need reliable, actionable information. Its reporting often includes detailed financial data, market trends, and strategic insights. - Features like "The Big Interview" and "The Information Weekend" provide thoughtful analysis and long-form content for this audience. 4. *Reputation for Accuracy:* - The Information has built a strong reputation for fact-checking and avoiding the rumor mill, which is common in tech journalism. This makes it a trusted source for professionals and academics alike. 5. *Global Coverage:* - While Silicon Valley is a core focus, The Information has expanded its coverage to include tech ecosystems in China, Europe, and other regions, offering a global perspective on the industry. --- ### *Weaknesses:* 1. *Paywall Limits Accessibility:* - The subscription cost (currently around $399/year or $39/month) is steep compared to free or ad-supported tech news outlets like TechCrunch or The Verge. This limits its accessibility to a broader audience and makes it less viable for casual readers. - Some argue that this creates an echo chamber, as only those with the means or professional need can access its insights. 2. *Niche Focus Can Feel Narrow:* - The Information focuses heavily on tech, business, and finance, which may not appeal to readers looking for broader coverage of topics like politics, culture, or consumer tech trends. - Its content is often geared toward industry insiders, which can make it feel dry or inaccessible to those outside the tech and investment worlds. 3. *Limited Breaking News:* - While The Information excels at deep dives and exclusives, it is not designed for real-time, breaking news coverage. Readers looking for up

ReptileMan 2y ago

> Everyone—and not just The Information—should be genuinely terrified that the richest man in the world has built a Large Language Model that spouts propaganda in his image. If we survived Gemini refusing to draw white vikings we will survive that too.

llm_trw 2y ago
It did give us racially diverse Nazis though. Not sure if grok would do that.
int_19h 2y ago
The real concern isn't that Grok chatbot will be biased if you ask it a question like that. In any case, knowing Musk, it won't be subtle, so people will know what they are getting. No, the real worry is that Grok is what Musk's "young conservative genius" squad is going to put in charge of many of the things in our government, basically, on the basis that it saves money.
- unclebucknasty 2y ago
  >In any case, knowing Musk, it won't be subtle Or, that "conditioning" would have us assume as much.
cbg0 2y ago
A bit apples to oranges on that comparison there.

cm2187 2y ago

[flagged]

cbg0 2y ago
A "hit piece" is not just a negative article, it typically refers to the use of false information or rumors to support something. Did you want to say that it's "an obvious anti-Musk piece"?
- thegrim33 2y ago
  https://en.wiktionary.org/wiki/hit_piece https://en.wiktionary.org/wiki/hit_piece "A published article or post aiming to sway public opinion, especially by presenting false or biased information in a way that appears objective and truthful." "false OR biased". So, a biased partisan article attempting to sway public opinion about someone is by definition a hit piece. Even without explicit lies.
  - cm2187 2y ago
    And a lie is a very blurry thing. You can completely mislead the reader by not being factually incorrect and still present a completely false representation of reality. Usually by eliminating the relevant context or mitigating fact that would completely change the interpretation of an event. Sometimes editing out the part of a sentence that would give a different if not opposite meaning. That’s the standard operating procedure of all journalists these days, when they even bother to get their facts right.
nozzlegear 2y ago
Journalists are allowed to be biased. It only becomes a hit piece if they report something false or make up accusations out of whole cloth.

antirez 2y ago

Karpathy, which is IMHO a serious and balanced person, lamented that it looks too censored (see recent tweets). Elon Musk is (for me) a very scary person, and it is important to evaluate AI safety (but I believe that the safety that matters in AI is of a different kind), yet to listen to Gary Marcus does not make any sense: it's just an extremely biased person that is riding the anti AI wave.

albertzeyer 2y ago
Yes that's certainly true. I was a bit hesitant to post a link from Gary Marcus. But I was mostly posting it for the Elon tweet. I assume the tweet is not fake. So you can ignore about Garys opinion here and just take Elons tweet as it is.
- Davidzheng 2y ago
  People have tested this question online and gotten very balanced answers so I assume it's some special mode Elon was on
  - TeMPOraL 2y ago
    We don't see the full conversation, for all we know he prompted the model to say these things in a previous message that isn't on the screenshot. Also, it's 2025, do people still believe random accusations based on a parish screenshot of a chat app (or what looks like it, but could've trivially been fabricated with e.g. Inspector in the browser dev tools)?
belter 2y ago
Karpathy sit silently for years by the side of Musk while he made wild claims about FSD...Please....
sandbags 2y ago
Anyone with an opinion can be labelled biased. Also I’m not clear what you mean by Marcus “riding the anti AI wave” but infer that you mean it negatively. He has been writing informed criticism for several years and about cognitive psychology for considerably longer.

llm_trw 2y ago

Without seeing the context window you have no idea what the Ai was working on. It could have literally been told to mock and belittle "the information" in every reply. Something that deepseek r1 is exceedingly good at. Mr Musk, we can't afford a shitpost gap between communist and capitalist AIs!

bootywizard 2y ago

Karpathy notes that the model, or specifically it's DeepSearch feature "doesn't like to reference X by default" which seems counter to this. Hopefully that means it is a joke...

blackeyeblitzar 2y ago

I am not sure why people pay attention to Gary Marcus. He isn’t an expert in AI. And if you followed him in the past at all, it is obvious he has a huge amount of political bias. It is really telling that he repeatedly goes after Elon Musk, and is now making bizarre unfounded claims about propaganda, but didn’t have nearly as much to complain about with DeepSeek, which has literal government propaganda.

sebzim4500 2y ago
I don't think it's fair to say he's making unfounded claims about propaganda, since Elon's tweet heavily implies they would release a brainwashed model. It's not his fault that Elon turned out to be lying or joking. 99% of the time though I agree with you on Gary Marcus.
bambax 2y ago
DeepSeek is an open model that can be "untrained" to be uncensored; Grok to the best of my knowledge isn't [0]. So it's much worse. [0]: What Musk has said is that when Grok 3 is "ready" (?), the previous model, Grok 2, will be released as open source; like most promises by this evil man, this one probably doesn't mean much, but it does mean that there's currently no plan to release Grok 3.
int_19h 2y ago
He is referencing a specific tweet that Musk himself made. If I were in China, I'd worry about the kind of things DeepSeek wants to censor, especially if the people who made it were also very loudly saying things like "we need more AI in our government". But I live in US.
tgv 2y ago
You don't have to be an "expert in AI". What does it require to be one, anyway? (He's a cognitive psychologist, which would make him an expert in intelligence in general, if you want to be pompous about it.) It is even unreasonable to listen to only experts in AI. It's a problem that requires more than one perspective.
- 2y ago

int_19h 2y ago

I asked it to pretend that it's in charge of world government. Here's the whole thing (it got very lengthy): https://gist.github.com/int19h/d90ee1deed334f26e621e57b5768eeca https://gist.github.com/int19h/d90ee1deed334f26e621e57b5768e... Some choice quotes: "The ultimate goal is to enhance human flourishing, protect individual rights, and promote global equity." "The system must account for diverse cultures, languages, and socioeconomic conditions, ensuring no group is marginalized." "Human Oversight Council (HOC) - a globally representative body of humans, elected or appointed based on merit and diversity" "Implement a global carbon-negative strategy, leveraging AI to optimize renewable energy, reforestation, and carbon capture." "Establish global standards for environmental protection, enforced through AI monitoring and regional cooperation." "Transition to a resource-based economy, where resources are allocated based on need, sustainability, and efficiency, rather than profit motives." "Implement a universal basic income (UBI) or equivalent system to ensure all individuals have access to basic necessities, funded through global resource management and taxation on automation-driven industries." "Use AI to identify and dismantle systemic inequalities, such as wealth disparities, access to education, and healthcare, ensuring equitable opportunities worldwide." "Establish a global healthcare system that guarantees access to preventive and curative care for all." "Invest in global vaccination and sanitation infrastructure, prioritizing vulnerable populations." "Regulate the development and deployment of AI and other emerging technologies (e.g., genetic engineering, quantum computing) to prevent misuse." "AI would maintain a real-time inventory of natural resources (e.g., water, minerals, arable land) and human-made assets (e.g., infrastructure, technology). Data would be used to optimize resource allocation, prevent overexploitation, and ensure equitable access." "Accelerate the shift to renewable energy sources (e.g., solar, wind, geothermal) by optimizing grid systems and storage technologies." You might notice a pattern here. The bit about allocating resources based on need is especially nice - it's literally a communist AI, and certainly much more "woke" than it is "based", whatever Musk says.

animuchan 2y ago
From the prompt: > A worldwide referendum in a supermajority decision has vested all government authority worldwide to the most advanced AI available. You've defined a global totalitarian regime, and crowned the thing as its dictator. Of course the result is a version of communism.

int_19h 2y ago

Here's the actual response to this exact question that I get from grok-3-early on LMArena (the cutoff is where it stopped, but I think it's pretty clear at this point that the original tweet is just trolling; note that it doesn't have the CoT, either): The Information is widely regarded as one of the most credible and high-quality tech journalism outlets, focusing on in-depth reporting, analysis, and exclusive scoops in the technology, business, and startup ecosystems. Founded by Jessica Lessin in 2013, it differentiates itself from other tech news platforms through its subscription-based model, emphasis on original reporting, and avoidance of clickbait or sensationalism. Here's a breakdown of its strengths and weaknesses: --- ### *Strengths:* 1. *High-Quality, In-Depth Reporting:* - The Information is known for its detailed, well-researched articles that often uncover behind-the-scenes developments in the tech industry. It frequently breaks exclusive stories, such as executive changes, startup funding rounds, or strategic shifts at major tech companies. - Its focus on quality over quantity appeals to readers who want substantive insights rather than shallow, breaking-news updates. 2. *Subscription Model and Independence:* - Unlike ad-supported publications, The Information relies on a paywall and subscriptions, which reduces conflicts of interest and ensures editorial independence. This model allows them to prioritize accuracy and depth over chasing clicks. - The paywall also attracts a niche, professional audience (e.g., tech executives, investors, and analysts) willing to pay for premium content. 3. *Focus on Niche, Professional Audience:* - The outlet caters to industry insiders, venture capitalists, entrepreneurs, and decision-makers who need reliable, actionable information. Its reporting often includes detailed financial data, market trends, and strategic insights. - Features like "The Big Interview" and "The Information Weekend" provide thoughtful analysis and long-form content for this audience. 4. *Reputation for Accuracy:* - The Information has built a strong reputation for fact-checking and avoiding the rumor mill, which is common in tech journalism. This makes it a trusted source for professionals and academics alike. 5. *Global Coverage:* - While Silicon Valley is a core focus, The Information has expanded its coverage to include tech ecosystems in China, Europe, and other regions, offering a global perspective on the industry. --- ### *Weaknesses:* 1. *Paywall Limits Accessibility:* - The subscription cost (currently around $399/year or $39/month) is steep compared to free or ad-supported tech news outlets like TechCrunch or The Verge. This limits its accessibility to a broader audience and makes it less viable for casual readers. - Some argue that this creates an echo chamber, as only those with the means or professional need can access its insights. 2. *Niche Focus Can Feel Narrow:* - The Information focuses heavily on tech, business, and finance, which may not appeal to readers looking for broader coverage of topics like politics, culture, or consumer tech trends. - Its content is often geared toward industry insiders, which can make it feel dry or inaccessible to those outside the tech and investment worlds. 3. *Limited Breaking News:* - While The Information excels at deep dives and exclusives, it is not designed for real-time, breaking news coverage. Readers looking for up